Article https://doi.org/10.

1038/s41586-018-0832-5

GABAA receptor signalling mechanisms

revealed by structural pharmacology
Simonas Masiulis1*, Rooma Desai2, Tomasz Uchański3,4, Itziar Serna Martin5, Duncan Laverty1, Dimple Karia5,
Tomas Malinauskas5, Jasenko Zivanov1, Els Pardon3,4, Abhay Kotecha6, Jan Steyaert3,4, Keith W. Miller2* & A. Radu Aricescu1,5*

Type-A γ-aminobutyric (GABAA) receptors are ligand-gated chloride channels with a very rich pharmacology. Some of
their modulators, including benzodiazepines and general anaesthetics, are among the most successful drugs in clinical
use and are common substances of abuse. Without reliable structural data, the mechanistic basis for the pharmacological
modulation of GABAA receptors remains largely unknown. Here we report several high-resolution cryo-electron
microscopy structures in which the full-length human α1β3γ2L GABAA receptor in lipid nanodiscs is bound to the
channel-blocker picrotoxin, the competitive antagonist bicuculline, the agonist GABA (γ-aminobutyric acid), and the
classical benzodiazepines alprazolam and diazepam. We describe the binding modes and mechanistic effects of these
ligands, the closed and desensitized states of the GABAA receptor gating cycle, and the basis for allosteric coupling
between the extracellular, agonist-binding region and the transmembrane, pore-forming region. This work provides a
structural framework in which to integrate previous physiology and pharmacology research and a rational basis for the
development of GABAA receptor modulators.

In vertebrates, GABAA receptors mediate both phasic and tonic neuronal
inhibition in the adult central nervous system1–3. Their dysfunction
leads to channelopathies associated with epilepsy, insomnia, anxiety
and chronic pain4. GABAA receptors are among the most important
human drug targets owing to their many allosteric sites, which bind
compounds with anticonvulsant, anti-anxiety, analgesic, sedative and
anaesthetic properties5,6. Some of these—such as benzodiazepines
(BZDs), which are typically positive allosteric modulators—entered
clinical use decades before the identity of their receptors was known7,8,
and were crucial for the isolation9 and subsequent cloning10 of GABAA
receptors. Other GABAA receptor ligands are important research tools,
including the antagonists picrotoxin (PTX) and bicuculline (BCC)5,6.
The binding modes and conformational effect of most GABAA
receptor allosteric modulators are unknown. Docking attempts have
had to rely on models based on the structures of homologous pro-
teins, including the Caenorhabditis elegans glutamate-gated chloride
channel α (GluCl)11, the Torpedo nicotinic acetylcholine receptor12 or
the human homopentameric β3 GABAA receptor13. Crystallographic
and single-particle cryo-electron microscopy (cryo-EM) studies have
revealed the interactions of GABAA-receptor-derived constructs with
neurosteroids14–16, the agonist GABA17–19, and the BZD-site ligands
bretazenil (BRZ)20 and flumazenil (FLZ)18,20. However, owing to the use
of engineered receptors, some of which are structurally damaged by the
presence of detergents, the interpretability of these models is limited.
Functional studies on small-molecule GABAA receptor modulators
have also raised numerous questions. We will list here just a few examples.
It is not known why the two agonist (GABA) binding sites, which
should be structurally identical, are not functionally equivalent21. The
mechanism of action of BZDs is also unclear, despite their widespread
use as sedatives and anxiolytics. Moreover—unlike newer compounds
such as bretazenil and flumazenil—classical BZDs including diaze-
pam (DZP, also known by the trade name Valium) and alprazolam
(ALP, also known by the trade name Xanax) act specifically through
GABAA receptors containing α1, α2, α3 or α5, but not α4 or α6 sub-
unit types22,23. The basis for this specificity is not known. The plant
alkaloid PTX, which is one of the most widely used GABAA receptor
antagonists24, is thought to act as a channel blocker. However, its com-
petitive antagonistic properties suggest that additional binding sites
might exist25,26. Another broadly used GABAA receptor antagonist,
BCC, is thought to act competitively, although its preference for bind-
ing resting over desensitized receptors suggests that it might also act
allosterically27. How reagents such as PTX and BCC actually bind, as
well as how they work, is not yet known.
To address these questions, here we present five structures of the
human synaptic α1β3γ2L GABAA receptor in complex with PTX, PTX
and GABA (PTX/GABA), BCC, DZP and GABA (DZP/GABA), and
ALP and GABA (ALP/GABA). We establish the binding modes and
structural effect of these ligands and explain the molecular basis for
their function. To obtain structures in which GABAA receptors can
be observed in physiologically relevant conformations, we used a full-
length receptor variant from a thoroughly characterized cell line28 and
reconstituted it in a lipid bilayer29. Our results lay the foundation for
understanding the fundamental principles of the action of small mol-
ecules on heteromeric synaptic GABAA receptors.
Picrotoxin and GABA binding modes
We first solved the structure of the α1β3γ2L receptor in complex with
PTX (800 μM) to a nominal resolution of 3.1 Å (Fig. 1a, Extended Data
Fig. 1a–f, Extended Data Table 1, Supplementary Video 1). PTX is an
equimolar mix of two highly similar compounds, picrotin and picro-
toxinin. Our structure illustrates picrotoxinin, which is known to be the
more active compound, fully sequestered in the channel pore between
the M2 2ʹ and 9ʹ rings (Fig. 1b, c, Extended Data Fig. 1g, Supplementary
Video 2). This agrees both with mutagenesis and electrophysiology

1
MRC Laboratory of Molecular Biology, Cambridge Biomedical Campus, Cambridge, UK. 2Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Harvard
Medical School, Boston, MA, USA. 3Structural Biology Brussels, Vrije Universiteit Brussel (VUB), Brussels, Belgium. 4VIB-VUB Center for Structural Biology, VIB, Brussels, Belgium. 5Division of
Structural Biology, Wellcome Centre for Human Genetics, University of Oxford, Oxford, UK. 6Materials and Structural Analysis, Thermo Fisher Scientific, Eindhoven, The Netherlands.
*e-mail: simonasm@mrc-lmb.cam.ac.uk; kwmiller@mgh.harvard.edu; radu@mrc-lmb.cam.ac.uk

N A T U R E | www.nature.com/nature
© 2018 Springer Nature Limited. All rights reserved.
 RESEARCH Article

a α1-D γ2-C b c
β3-E M2 M2 M2
N-linked γ2-C
Mb38 Loop-C β3-B M2
glycans
α1-A L274
L259, 9′
L274, 9′
M2 T271
L264, 9′
β3-B 72°
T271, 6′ 90°
T256, 6′ M2
β3-E Out
T261, 6′ T261

90° α1-D
L259
A252, 2′

TMD
PTX M2
T256
Loop-C γ2-C V257, 2′
α1-D S267, 2′ L264 M2
In α1-A β3-E

ICD
PIP2 β3-B γ2-C α1-D

Fig. 1 | Structure of the α1β3γ2L GABAA receptor in complex with down (c) views of PTX (carbon atoms in pink, oxygen atoms in red) bound
picrotoxin. a, Cryo-EM map of the PTX-bound α1β3γ2L GABAA receptor to the channel pore, with the amino acid side chains lining the site shown
viewed from the extracellular space (left) and parallel to the membrane as sticks. Dashed lines indicate hydrogen bonds.
plane (right). The PTX-binding site is boxed. b, c, Side-on (b) and top-

results that implicate the 2ʹ, 6ʹ and 9′ M2 residues in PTX binding30–32,
and with suggestions that PTX becomes ‘trapped’ in closed/rest-
ing channels of GABAA and Gly receptors upon agonist wash-off33,34.
The hydrophobic isoprenyl moiety is surrounded by the 9ʹ Leu ring,
whereas the exocyclic oxygen atoms in the main PTX body form puta-
tive hydrogen bonds with the 6ʹ ring (Fig. 1b, c). In picrotin, the isoprenyl
group is replaced by a polar tertiary alcohol; this is not compatible
with 9′ Leu ring interactions, which explains why picrotin is less active.
We next determined a GABAA receptor structure in complex with
PTX (800 μM) and the neurotransmitter GABA (5 μM) to a nominal
resolution of 3.04 Å (Fig. 2a, Extended Data Fig. 2a–c, Extended Data
Table 1, Supplementary Video 3). The orthosteric GABA-binding
sites are located at the two β3+/α1− interfaces (the principal (+) and
complementary (−) annotation of subunit faces is used here), capped
by the extracellular loops-C of the β subunits18,19,29. Similar to the
Torpedo nAChR α subunits35, in the agonist-free GABAA receptor
the β3 loops-C adopt ‘open’ (outwards-projecting) conformations,
whereas in the PTX/GABA-bound state they are ‘closed’ (Extended
Data Fig. 2d). Strong electron microscopy densities are present in the
two orthosteric ligand-binding sites, enabling unambiguous model-
ling of GABA molecules in the ‘aromatic boxes’ formed by β3Tyr157,
β3Phe200, β3Tyr205 and α1Phe65 (Fig. 2b, Extended Data Fig. 2e,
Supplementary Video 4). The GABA amino group engages in a cation–π
interaction with β3Tyr205 and a network of hydrogen bonds

a α1-D γ2-C
b c
Mb38 β3-E
Loop-C 35
Loop-C
α1-A
30
Distance along pore (Å)

Y157
E155 25
β3-B Y97
Y205 F65 20
β6 9′
GABA
T130
β3-E F200 15 PTX 6′
β2 10
90° 2′ PTX
T202
TMD

Loop-C 5 PTX/
γ2-C PTX –2′ GABA
R67
β3-E+ α1-D– β1 0
α1-D 0 1 2 3 4 5
Pore radius (Å)
PIP2

d e α1 helix f α1 helix
GABA
1.3° β3-B+ α1-A– β3-E+ α1-D–

2.6° ECD rotation F15

F31

γ2-C F15 β3 β5 F31 β3

β3-B β5
Y159 Y159
R85 β6
D163 β6 D163
R85
α1-D
α1-A 0.9° R120 R120
β2
β3-E
β2 Loop-B
Loop-B

β1
1.3° Loop-C Y205 GABA
Y205 GABA
β1
PTX
Loop-C Loop-C
PTX/GABA 4.5°

Fig. 2 | Conformational impact of GABA binding to the α1β3γ2L
GABAA receptor. a, Cryo-EM map of the PTX/GABA-bound α1β3γ2L
receptor viewed from the extracellular space (left) and parallel to the
membrane plane (right). b, One GABA-binding pocket viewed from the
extracellular space. GABA is shown in ball-and-stick representation with
the atoms coloured as follows: carbon, khaki; oxygen, red; nitrogen, blue.
c, Plot of the pore radii for the receptor bound to PTX (black dashed
line) and PTX/GABA (green line). d, Superposition of ECDs from the
PTX-bound (grey) and the PTX/GABA-bound (red/blue/yellow) receptor
structures based on the global TMD alignment. Subunits were radially
translated away by 10 Å from the pore axis to enable better visualization of
conformational changes in the ECD upon GABA binding. GABA-induced
ECD rotation is stated as the rotation angle around the ECD rotation axes,
and the direction of motion is shown. e, f, Superposition of α1 ECDs from
PTX-bound and PTX/GABA-bound structures reveals conformational
changes induced by GABA binding in the orthosteric pockets at β3-B+/
α1-A− (e) and β3-E+/α1-D− (f) interfaces. Dashed lines indicate hydrogen
bonds, π–π stacking, π–cation interactions and salt bridges.

N A T U R E | www.nature.com/nature
© 2018 Springer Nature Limited. All rights reserved.
 Article RESEARCH

a 3 μM Mb38
b 3 μM Mb38 100 μM BCC c 3 μM Mb38
of the extracellular domain (ECD) and consequently modulates agonist
binding. Indeed, PTX has been shown to reduce the apparent agonist
affinity in GABAA receptors26,36 and to increase GABA dissociation37.
Therefore, TMD closure by PTX constitutes the basis for allosteric com-
munication between the PTX and GABA binding sites and explains
how PTX appears to act as a competitive inhibitor34,38,39 without
binding to other receptor pockets. We conclude that the PTX-bound
200 pA
10 s structure best illustrates the α1β3γ2L receptor in a closed/resting state
d (closed TMD, agonist-free ECD).
α1-A β3-E α1-D γ2-C
β3-E
β3-B
Mechanistic impact of GABA binding
Previous observations suggest that the two GABA-binding sites
Loop-C
may not be functionally equivalent21,40, and the β-E+/α-D− site has
been estimated to have a threefold higher affinity for GABA than
BCC
the β-B+/α-A− site21. By comparing ECDs in the PTX-bound and
Mb38
the PTX/GABA-bound structures, both of which have closed state
Out TMDs, it is possible to visualize the initial conformational changes at
the two agonist-binding sites upon the binding of GABA. Closure of β3
140° loops-C—likely to be the first step in the process—triggers remodelling
of the β3-B+/α1-A− and the β3-E+/α1-D− interfaces, which leads to
an anticlockwise (looking down the pore axis from the extracellular
In space) asymmetric rotation of all subunit ECDs (Fig. 2d). This state is
PIP2 stabilized by new hydrogen-bond networks involving the β3Gly158,
e f β3Tyr205 and α1Arg85 residues. The cation–π interaction between
35 α1Arg85 and β3Tyr159 breaks, and the α1Arg85 side chain moves to
β5
30 the periphery of the interface (Fig. 2e, f). At the β3-B+/α1-A− inter-
Loop-B
Distance along pore (Å)

β6
Y205
Y157
R67
β2
Loop-C
β1
F200
β3-B+
F46
α1-A–
Fig. 3 | Structure of a α1β3γ2L GABAA receptor closed by the
competitive antagonist bicuculline. a–c, Representative whole-cell
current traces elicited from the same cell (n = 4) by a 42-s pulse of: Mb38
alone (a); Mb38 and BCC co-applied for 13 s at the 25-s mark (b); Mb38
again, after BCC was washed out (c). d, Cryo-EM map of the α1β3γ2
GABAA receptor–BCC complex viewed parallel to the membrane plane.
e, One BCC binding pocket at the β3+/α1− interface, viewed parallel to the
membrane. Dashed lines indicate π–π interactions and hydrogen bonds.
f, Plot of the pore radii for the α1β3γ2 receptor bound to PTX (black
dashed line) and BCC (magenta line).
involving the β3Glu155 carboxyl, main-chain carbonyls of β3Ser156
and β3Tyr157, and the β3Tyr97 hydroxyl. The GABA carboxylate forms
salt bridges with α1Arg67 and hydrogen bonds with α1Thr130 and
β3Thr202 (Fig. 2b). The GABA binding modes that we observe are
generally consistent with those proposed in lower-resolution α1β1γ2
and α1β2γ2 GABAA receptor models18,19; however, there are some dif-
ferences in the details, illustrated in the accompanying Letter29.
Allosteric cross-talk between PTX and GABA
Co-application of PTX (800 μM) with GABA (5 μM) and megabody 38
(Mb38; 2 μM; ref. 29) causes rapid and complete inhibition of whole-cell
currents in HEK293 cells, with complete recovery after PTX washout
(Extended Data Fig. 2f–h). Unexpectedly, in both PTX-bound and
PTX/GABA-bound structures, the transmembrane domains (TMDs)
adopt the same conformation—all five M2 helices present the hydro-
phobic 9′ Leu side chains towards the centre of the pore, constricting the
pore radius to ~1.5 Å (Fig. 2c, Extended Data Fig. 2i). This suggests that
PTX inhibition is not caused by blocking an open pore, but by inducing
and maintaining a closed pore conformation. It is likely that, by sta-
bilizing the closed/resting TMD state, PTX affects the conformation
face, these changes enable the β3-B subunit to move closer to the α1-A
25
subunit by 1.5 Å (as measured between β3Gly33Cα and α1Pro80Cα),
20 9′ and the total buried surface between the two subunit ECDs increases
15 by 184 Å2 (Extended Data Fig. 2j, Extended Data Table 2). In the
GABA-bound conformation, α1Phe15 and β3Phe31 residues form a
10 2′ PTX hydrogen–π interaction and the side chains of α1Arg85 and β3Asp163
5 –2′
BCC form new salt bridges, further interlocking the β3-E+/α1-D− ECD
interface and increasing its total buried surface area by 234 Å2 (Fig. 2f,
0
0 1 2 3 4 5 Extended Data Table 2, Supplementary Video 5).
Pore radius (Å) The ‘incomplete’ closure of the β3-B+/α1-A− interface in the PTX/
GABA-bound structure, relative to the β3-E+/α1-D− interface, defines
an intermediate conformation from which GABA can presumably
dissociate more readily. Notably, in the desensitized (high affinity to
GABA41) receptor states—described later in this paper—both β3+/
α1− interfaces close to the same extent. We conclude that the PTX/
GABA-bound structure represents a pre-active receptor state, in which
agonist-induced conformational changes at the ECD level are not large
enough to perturb the closed/resting TMD conformation42.
Mechanism of bicuculline antagonism
Being a highly efficient competitive antagonist, BCC was expected
to induce a true GABAA receptor closed state43. We first verified
whether BCC could close the α1β3γ2L receptor that had been pre-
bound to Mb38. In whole-cell voltage-clamp experiments, applica-
tion of a 42-s pulse of Mb38 produced currents (Fig. 3a) that were
inhibited (102 ± 7%; n = 4) by the co-application of bicuculline (100
μM) (Fig. 3b) in a reversible fashion (Fig. 3c). We solved the cryo-EM
structure of the α1β3γ2L heteromer in complex with Mb38 and BCC
to a nominal resolution of 3.69 Å (Fig. 3d, Extended Data Fig. 3a–c,
Extended Data Table 1). Electron microscopy densities corresponding
to BCC were observed at both of the orthosteric agonist sites (Extended
Data Fig. 3d), where the hydrophobic nature of the phthalide
and isoquinoline rings of BCC enable its interaction with ‘aromatic
box’ residues β3Tyr157, β3Phe200, β3Tyr205 and α1Phe65 (Fig. 3e,
Supplementary Video 6). Relative to the PTX-bound structure, the
β3-B and β3-E subunit loops-C flex inward (by around 2.2 Å at the tip;
Extended Data Fig. 3e) in order to accommodate BCC, but they retain
overall ‘open’ conformations (Extended Data Fig. 3f). The channel
pore of the BCC-bound α1β3γ2L receptor is fully closed by the M2
residues at three levels—9ʹ, 2ʹ and −2ʹ (Fig. 3f)—and its subunit con-
formations are almost identical to those of the PTX-bound structure

N A T U R E | www.nature.com/nature
© 2018 Springer Nature Limited. All rights reserved.
 RESEARCH Article
a Diazepam b α1-D γ2-C
3 O
N β3-E
N
5 1
6 A 9 β6
7 8
Cl
Alprazolam
4 N Out
N N
I Y58
N F77
6 1
7 10
8 9 In
Cl PIP2
e f
γ2-C–
Loop-C
Y210 M2–M3 loop
β6
S205 M3
β2 M286
F100
DZP β1
Y58
F77
H102
F289
α1-D+ β3-E+
N60
Fig. 4 | Structures of a α1β3γ2L GABAA receptor in desensitized states
induced by GABA and alprazolam or GABA and diazepam.
a, Structural formulae of diazepam and alprazolam. Diazepine ring atoms
are numbered. Imidazole (I) and benzene (A) rings are labelled. b, c, The
cryo-EM map of the α1β3γ2 GABAA receptor in complex with ALP (cyan)
(b) and DZP (teal) (c) viewed parallel to the membrane plane. d, e, Views
of the benzodiazepine binding site at the α1+/γ2− interface showing ALP
(d) and DZP (e) binding modes. Dashed lines indicate π–π interactions
and hydrogen bonds. f, The low-affinity DZP-binding site in the β3+/
(Extended Data Fig. 3e, f). The binding of BCC at the orthosteric sites
prevents closure of the β3+/α1− subunit interfaces and rotation of the
ECDs, and also stabilizes TMDs in the closed state, thus inactivating
the channel.
Benzodiazepine binding modes and mechanisms
We next solved cryo-EM structures of the α1β3γ2L receptor in com-
plex with GABA and ALP, as well as with GABA and DZP, to nominal
resolutions of 3.26 Å and 3.58 Å, respectively (Fig. 4a–c, Extended Data
Fig. 4a–f, Extended Data Table 1, Supplementary Video 7). In both
structures, GABA molecules are bound to the orthosteric agonist pock-
ets and ALP or DZP occupy the canonical BZD-binding site at the α1+/
γ2− interface, where they form extensive interactions (Fig. 4d, e). The
densities of ALP and DZP were well defined, enabling us to distinguish
the fused benzene-diazepine from the pendant phenyl rings (Extended
Data Fig. 4g, h, Supplementary Video 8).
The binding modes of ALP and DZP shown here are in agreement
with the results of cysteine crosslinking experiments, in which isothi-
ocyanate substitutions at the DZP C7 position reacted with cysteines
introduced at the α1His102, α1Asn103 and γ2Asn60 positions, and
isothiocyanate introduced at DZP C3 reacted with the α1Ser206Cys
and α1Thr207Cys mutants44. The chlorine atoms at the C8 and C7
positions in ALP and DZP, respectively, interact with the α1His102 side
chain. In the α4 and α6 subunits the equivalent positions are occupied
by arginine residues, the larger side chains of which would sterically
clash with ALP and DZP. This could explain why classical BZDs do not
act on GABAA receptor subtypes containing α4 or α6 subunits22,45,46.
The ALP-bound and DZP-bound structures also demonstrate that
BZDs with a pendant phenyl ring do not share the binding mode
reported for the benzodiazepine antagonist FLZ and the partial ago-
nist BRZ18,20 (Extended Data Fig. 5a–e). Compared with ALP and DZP,
both FLZ and BRZ bind deeper into the BZD pocket and higher from
the pocket floor, which is delineated by the side chain of γ2Asn60. This
could be due to additional hydrogen bonds between the hydroxyl group
of γ2Thr142 and either the imidazole nitrogen or the ester carbonyl of
FLZ and BRZ20 (Extended Data Fig. 5c, e).
N A T U R E | www.nature.com/nature
© 2018 Springer Nature Limited. All rights reserved.
c α1-D γ2-C d
β3-E γ2-C–
Loop-C
Y210
S205
β2
ALP DZP F100
ALP β1
H102
α1-D+ N60
g h
α1-D– 35
GABA + Mb38 + DZP
30
Distance along pore (Å)
M2 25
N265 20
9′
DZP 15 100 pA
M1 10 s
PTX
10
ALP/GABA
5 DZP/GABA
–2′
0
M236 0 1 2 3 4 5
Pore radius (Å)
α1− interface TMD region. g, Plot of the pore radii for the receptor
bound to PTX (black dashed line), ALP/GABA (blue line) and DZP/
GABA (red dashed line). h, Representative current traces evoked by co-
application of GABA (10 mM) with Mb38 (2 μM) and DZP (100 μM) for
40 s to outside-out patches pulled from HEK293 cells. The currents (n = 6
patches) desensitized completely in three phases: at slow (0.78 ± 0.84 s−1),
medium (16.32 ± 9.34 s−1) and fast (306.5 ± 185.9 s−1) rates. Rate values
are mean ± s.d. The upper black solid bar shows the duration of ligand
application.
In the DZP/GABA-bound map, we also observed two strong densities
at sites in the β3+/α1− TMD interfaces that are thought to be the binding
sites of general anaesthethics47 (Extended Data Fig. 4i, Supplementary
Video 9). These densities are not present in the ALP/GABA-bound map
and have a distinct shape that is attributable to DZP. In these pockets,
the DZP A ring forms hydrophobic interactions with β3Met286 and
β3Phe289 and the phenyl ring points towards α1Pro233 (Fig. 4f, Extended
Data Fig. 4i). Our structure corroborates previous electrophysiology and
mutagenesis data, which suggests the existence of secondary DZP-binding
sites that are responsible for the anaesthetic activity and biphasic GABAA
receptor potentiation at higher DZP concentrations40,48.
We next compared the α1+/γ2− interfaces in PTX-, PTX/GABA-,
BCC-, ALP/GABA- and DZP/GABA-bound structures. Notably, the
binding of ALP or DZP induced only minor rearrangements in the
BZD pocket: γ2Asn60 adopted a different rotamer (Extended Data
Fig. 5f) and the tip of the α1 loop-C flexed outwards by around 0.6
Å (Extended Data Fig. 5f, g). We therefore propose that BZDs such as
ALP and DZP act as ‘connectors’ to stabilize the weakest ECD interface,
α1-D+/γ2-C− (Extended Data Table 2), and facilitate the concerted
ECD rotation upon GABA binding. Nanobody 38 (Nb38) and Mb38
might act similarly by ‘crosslinking’ the α1-A+/β3-E− and α1-D+/
γ2-C− interfaces20,29. FLZ antagonizes classical BZDs by competing for
the same pocket; however, because it interacts largely with the α1-D+
face18,20, its binding would not confer the same inter-subunit connec-
tivity benefits (Extended Data Fig. 5b, c).
The channel-pore radii in the ALP/GABA-bound and the DZP/
GABA-bound structures are 2.6 Å and 2.3 Å, respectively, at the 9′
level, and are further constricted to 1.6 Å and 1.8 Å by the −2′ res-
idues (Fig. 4g). Whole-cell voltage-clamp experiments indicate that
a prolonged exposure to a combination of GABA (10 mM), Mb38 (2
μM) and diazepam (100 μM) leads to the complete desensitization of
the α1β3γ2L receptor (Fig. 4h). Subunit comparison between ALP/
GABA-bound and DZP/GABA-bound structures reveals that they
adopt similar conformations (Extended Data Fig. 6); this illustrates a
desensitized state of the receptor, in which it is bound to the agonist
but the channel is closed only at the −2′ gate (Supplementary Video 7).
 Article RESEARCH

a 6.3° GABA b 5.3° interfaces (Fig. 5b, Extended Data Fig. 7c). These motions are trans-
8.5° ALP mitted through interactions between the β1–β2 and the M2–M3 loops,
γ2-C
whereas the β6–β7 (Cys) loops act as pivot points and the TMD bundles
Loop-C
4.7°
1
rotate as rigid bodies (Extended Data Figs. 7d, 8a–f, Supplementary
α1-D
γ2-C 4 Video 10).
β3-B
β3-B 3 2 4.9° The relative flexibility in the M2–M3 loops appears to regulate the
α1-D 9′ Leu efficiency of the ECD–TMD signal transduction. In α1 and γ2 subu-
α1-A
β3-E
nits, the highly conserved arginine residues at the M2 19′ positions
4.2° interact with and rigidify the M2–M3 loops. However, in β3 subunits,
β3-E
5.2° the strictly conserved Lys279 residues displace the 19′ Arg269 side
α1-A
PTX Loop-C 5.1° 4.1° chains, causing them to rotate and wedge in between the M1 and M2
M2–M3 loop
ALP/GABA
8.9° helices of the neighbouring α1 subunits. This restricts the β3 TMD
c
rotations (Extended Data Fig. 8g–l), possibly dampening signal trans-
Closed/resting
Pull duction. In α1 and γ2 subunits, positions equivalent to β3Lys279 are
Pull
occupied by conserved threonine residues, which do not clash with 19′
γ Rotation γ Arg (Extended Data Fig. 8g–l). The M2–M3 loops of β3 therefore seem
β β to be more flexible than those of α1 and γ2 (Extended Data Fig. 7d,
α
GABA
α Supplementary Video 10). In agreement with these observations,
Lock α1β3γ2L receptors containing β3Lys279Thr mutations are 20-fold
α β
ALP
α β
Lock more sensitive to GABA relative to the wild-type receptors49.
We propose that GABA-induced signalling can be described by a
Rotation ‘lock-and-pull’ mechanism. GABA binding triggers loops-C closure in
d Pull β subunits, initiating rotation of their ECDs, which become ‘locked’ to
GABA the neighbouring α− interfaces. These conformational changes ‘pull’
the other ECDs, and transfer to the TMDs, leading to a concerted anti-
β α α γ γ clockwise rotation. The inclusion of a BZD strengthens the α1+/γ2−
β ECD interface to facilitate these motions (Fig. 5c). In β subunits, signal
V53 H56 I68 transduction to TMDs is modulated by the flexible M2–M3 loops,
P273 P278 P288
whereas the ECD rotations of α and γ subunits couple to the TMDs
3 2 9′ 3 2 L259 3 2 9′ 3 2 L264 3 2 9′ 3 2 L274 more efficiently because their M2–M3 loops are more rigid (Fig. 5d).
A253
–2′ A248 –2′ –2′ P263
Conclusion
The structures presented here illustrate how important pharmacological
Closed/ Desensitized compounds, used broadly in research and in the clinic, interact with a
resting
full-length human heteromeric GABAA receptor to modulate its confor-
Fig. 5 | Conformational differences between closed/resting and mation and function. Picrotoxin must initially bind to an open channel
desensitized states in a GABAA receptor. a, Superposition of ECDs from pore and subsequently stabilizes a closed/resting receptor state, which
PTX (grey) and ALP/GABA (red/blue/yellow) structures based on the explains its simultaneous channel blocker and allosteric antagonist activ-
global TMD alignment. Subunits were radially translated by 10 Å away ities. Bicuculline occupies the agonist-binding sites. However, unlike
from the pore axis to enable better visualization of conformational changes GABA, it cannot drive the rotation of β subunits and therefore stabilizes
in the ECD upon GABA and ALP binding. GABA- and ALP-induced
the closed channel pore. Comparison of agonist-free and GABA-bound
ECD rotation angles around the rotational ECD axes, and the direction of
motion are shown. b, Global TMD alignment for the PTX-bound (grey) structures delineates the molecular mechanism by which neurotrans-
and ALP/GABA-bound (coloured) structures. 9ʹ Leu side chains are shown mitter binding to the β3+/α1− interfaces prompts a global rotation of
as sticks. c, Schematic illustration of conformational changes initiated by ECD regions, initially in an asymmetrical fashion, and explains how
GABA binding at the ECD level. GABA stabilizes closure of loop-C in each different subunit types transduce this conformational change to their
β subunit, causing ECDs to rotate and form stronger β3+/α1− interfaces. TMDs. We also characterize the binding sites of two major classical
The black arrows indicate the direction of rotation, with increasing arrow benzodiazepines—alprazolam and diazepam—and define their pri-
thickness representing a greater magnitude of rotation. BZDs such as mary role as stabilizers of the α+/γ− interface, facilitating the concerted
alprazolam bind at the α1+/γ2− interface and reinforce it, facilitating motion of GABAA receptor subunits. This work underlines the poten-
the concerted rotation of the ECDs. Black bars (‘stitches’) at the subunit tial of cryo-EM to study the interactions of drugs with challenging yet
interfaces represent the strength of the interfaces (see Extended Data
Table 2). d, Differences in the relative orientations of ECD–TMD between
highly valuable human membrane-protein targets50. Specifically, these
the closed/resting and desensitized states illustrate how GABA binding structures might lead to the rational design of safer and more specific
and ECD rotation affect TMDs. Notably, the M2–M3 loops in β subunits anxiolytic, sedative, hypnotic and anticonvulsant drugs.
deform more than the α and γ equivalents, resulting in lower degrees of
M2 tilt and TMD rotation. Online content
Any methods, additional references, Nature Research reporting summaries, source
data, statements of data availability and associated accession codes are available at
Allosteric interactions between ECDs and TMDs https://doi.org/10.1038/s41586-018-0832-5.
To understand how the GABA-induced ECD rotation induces confor- Received: 17 August 2018; Accepted: 23 November 2018;
mational changes in the TMD region, we compared ALP/GABA-bound Published online xx xx xxxx.
and PTX-bound structures (Fig. 5a, b). In the presence of ALP, GABA
binding leads to equal closure of both β3+/α1− interfaces and a larger 1. Sieghart, W. Structure and pharmacology of γ-aminobutyric acidA receptor
ECD rotation compared to the pre-open PTX/GABA-bound structure subtypes. Pharmacol. Rev. 47, 181–234 (1995).
(Fig. 5a, Extended Data Figs. 2j, k, 7a, b, Supplementary Video 10). 2. Barnard, E. A. et al. International Union of Pharmacology. XV. Subtypes of
This conformational change of the ECD triggers an anticlockwise γ-aminobutyric acidA receptors: classification on the basis of subunit structure
and receptor function. Pharmacol. Rev. 50, 291–313 (1998).
rotation of the TMD. The resulting tilt of M2 helices moves the 9′ Leu 3. Sigel, E. & Steinmann, M. E. Structure, function, and modulation of GABAA
side chains away from the channel pore and towards the inter-subunit receptors. J. Biol. Chem. 287, 40224–40231 (2012).

N A T U R E | www.nature.com/nature
© 2018 Springer Nature Limited. All rights reserved.
 RESEARCH Article
4. Möhler, H. GABAA receptors in central nervous system disease: anxiety,
epilepsy, and insomnia. J. Recept. Signal Transduct. 26, 731–740 (2006).
5. Olsen, R. W. Analysis of γ-aminobutyric acid (GABA) type A receptor subtypes
using isosteric and allosteric ligands. Neurochem. Res. 39, 1924–1941 (2014).
6. Olsen, R. W. Allosteric ligands and their binding sites define γ-aminobutyric acid
(GABA) type A receptor subtypes. Adv. Pharmacol. 73, 167–202 (2015).
7. Hunkeler, W. Benzodiazepines, the story of the antagonist flumazenil and of the
partial agonist bretazenil. Chimia 47, 141–147 (1993).
8. Squires, R. F. & Brastrup, C. Benzodiazepine receptors in rat brain. Nature 266,
732–734 (1977).
9. Sigel, E., Mamalaki, C. & Barnard, E. A. Isolation of a GABA receptor from
bovine brain using a benzodiazepine affinity column. FEBS Lett. 147, 45–48
(1982).
10. Schofield, P. R. et al. Sequence and functional expression of the GABAA receptor
shows a ligand-gated receptor super-family. Nature 328, 221–227 (1987).
11. Hibbs, R. E. & Gouaux, E. Principles of activation and permeation in an
anion-selective Cys-loop receptor. Nature 474, 54–60 (2011).
12. Miyazawa, A., Fujiyoshi, Y. & Unwin, N. Structure and gating mechanism of the
acetylcholine receptor pore. Nature 423, 949–955 (2003).
13. Miller, P. S. & Aricescu, A. R. Crystal structure of a human GABAA receptor.
Nature 512, 270–275 (2014).
14. Miller, P. S. et al. Structural basis for GABAA receptor potentiation by
neurosteroids. Nat. Struct. Mol. Biol. 24, 986–992 (2017).
15. Laverty, D. et al. Crystal structures of a GABAA-receptor chimera reveal new
endogenous neurosteroid-binding sites. Nat. Struct. Mol. Biol. 24, 977–985
(2017).
16. Chen, Q. et al. Structural basis of neurosteroid anesthetic action on GABAA
receptors. Nat. Commun. 9, 3972 (2018).
17. Liu, S. et al. Cryo-EM structure of the human α5β3 GABAA receptor. Cell Res. 28,
958–961 (2018).
18. Zhu, S. et al. Structure of a human synaptic GABAA receptor. Nature 559, 67–72
(2018).
19. Phulera, S. et al. Cryo-EM structure of the benzodiazepine-sensitive α1β1γ2S
tri-heteromeric GABAA receptor in complex with GABA. eLife 7, e39383
(2018).
20. Miller, P. et al. Heteromeric GABAA receptor structures in positively-modulated
active states. Preprint at https://www.biorxiv.org/content/
early/2018/06/05/338343 (2018).
21. Baumann, S. W., Baur, R. & Sigel, E. Individual properties of the two functional
agonist sites in GABAA receptors. J. Neurosci. 23, 11158–11166 (2003).
22. Tan, K. R., Rudolph, U. & Lüscher, C. Hooked on benzodiazepines: GABAA
receptor subtypes and addiction. Trends Neurosci. 34, 188–197 (2011).
23. Knoflach, F. et al. Pharmacological modulation of the diazepam-insensitive
recombinant gamma-aminobutyric acidA receptors alpha 4 beta 2 gamma 2
and alpha 6 beta 2 gamma 2. Mol. Pharmacol. 50, 1253–1261 (1996).
24. Johnston, G. A. R. Advantages of an antagonist: bicuculline and other GABA
antagonists. Br. J. Pharmacol. 169, 328–336 (2013).
25. Krishek, B. J., Moss, S. J. & Smart, T. G. A functional comparison of the
antagonists bicuculline and picrotoxin at recombinant GABAA receptors.
Neuropharmacology 35, 1289–1298 (1996).
26. Goutman, J. D. & Calvo, D. J. Studies on the mechanisms of action of picrotoxin,
quercetin and pregnanolone at the GABAρ1 receptor. Br. J. Pharmacol. 141,
717–727 (2004).
27. Xu, X.-J., Roberts, D., Zhu, G.-N. & Chang, Y.-C. Competitive antagonists facilitate
the recovery from desensitization of α1β2γ2 GABAA receptors expressed in
Xenopus oocytes. Acta Pharmacol. Sin. 37, 1020–1030 (2016).
28. Dostalova, Z. et al. Human α1β3γ2L gamma-aminobutyric acid type A
receptors: high-level production and purification in a functional state. Protein
Sci. 23, 157–166 (2014).
29. Laverty, D. et al. Cryo-EM structure of the human α1β3γ2 GABAA receptor in a
lipid bilayer. Nature https://doi.org/10.1038/s41586-018-0833-4 (2019).
30. Chen, L., Durkin, K. A. & Casida, J. E. Structural model for gamma-aminobutyric
acid receptor noncompetitive antagonist binding: widely diverse structures fit
the same site. Proc. Natl Acad. Sci. USA 103, 5185–5190 (2006).
31. Xu, M., Covey, D. F. & Akabas, M. H. Interaction of picrotoxin with GABAA receptor
channel-lining residues probed in cysteine mutants. Biophys. J. 69, 1858–1867
(1995).
32. Gurley, D., Amin, J., Ross, P. C., Weiss, D. S. & White, G. Point mutations in the M2
region of the alpha, beta, or gamma subunit of the GABAA channel that abolish
block by picrotoxin. Recept. Channels 3, 13–20 (1995).
33. Bali, M. & Akabas, M. H. The location of a closed channel gate in the GABAA
receptor channel. J. Gen. Physiol. 129, 145–159 (2007).
34. Wang, D.-S., Mangin, J.-M., Moonen, G., Rigo, J.-M. & Legendre, P. Mechanisms
for picrotoxin block of α2homomeric glycine receptors. J. Biol. Chem. 281,
3841–3855 (2006).
35. Unwin, N. Nicotinic acetylcholine receptor and the structural basis of
neuromuscular transmission: insights from Torpedo postsynaptic membranes.
Q. Rev. Biophys. 46, 283–322 (2013).
36. Qian, H., Pan, Y., Zhu, Y. & Khalili, P. Picrotoxin accelerates relaxation of GABAC
receptors. Mol. Pharmacol. 67, 470–479 (2005).
37. Chang, Y. & Weiss, D. S. Site-specific fluorescence reveals distinct structural
changes with GABA receptor activation and antagonism. Nat. Neurosci. 5,
1163–1168 (2002).
N A T U R E | www.nature.com/nature
© 2018 Springer Nature Limited. All rights reserved.
38. Mehta, A. K. & Ticku, M. K. Characterization of the picrotoxin site of GABAA
receptors. Curr. Protoc. Pharmacol. 8, (2001).
39. Talwar, S. & Lynch, J. W. Phosphorylation mediated structural and functional
changes in pentameric ligand-gated ion channels: implications for drug
discovery. Int. J. Biochem. Cell Biol. 53, 218–223 (2014).
40. Olsen, R. W. GABAA receptor: Positive and negative allosteric modulators.
Neuropharmacology 136 (Pt A), 10–22 (2018).
41. Chang, Y., Ghansah, E., Chen, Y., Ye, J. & Weiss, D. S. Desensitization mechanism
of GABA receptors revealed by single oocyte binding and receptor function.
J. Neurosci. 22, 7982–7990 (2002).
42. Gielen, M. & Corringer, P.-J. The dual-gate model for pentameric ligand-gated
ion channels activation and desensitization. J. Physiol. (Lond.) 596, 1873–1902
(2018).
43. Ueno, S., Bracamontes, J., Zorumski, C., Weiss, D. S. & Steinbach, J. H.
Bicuculline and gabazine are allosteric inhibitors of channel opening of the
GABAA receptor. J. Neurosci. 17, 625–634 (1997).
44. Middendorp, S. J. et al. Relative positioning of classical benzodiazepines to the
γ2-subunit of GABAA receptors. ACS Chem. Biol. 9, 1846–1853 (2014).
45. Wieland, H. A., Lüddens, H. & Seeburg, P. H. A single histidine in GABAA
receptors is essential for benzodiazepine agonist binding. J. Biol. Chem. 267,
1426–1429 (1992).
46. Rudolph, U. & Knoflach, F. Beyond classical benzodiazepines: novel therapeutic
potential of GABAA receptor subtypes. Nat. Rev. Drug Discov. 10, 685–697
(2011).
47. Forman, S. A. & Miller, K. W. Anesthetic sites and allosteric mechanisms of
action on Cys-loop ligand-gated ion channels. Can. J. Anaesth. 58, 191–205
(2011).
48. Walters, R. J., Hadley, S. H., Morris, K. D. W. & Amin, J. Benzodiazepines act on
GABAA receptors via two distinct and separable mechanisms. Nat. Neurosci. 3,
1274–1281 (2000).
49. Fisher, J. L. A lysine residue in the β3 subunit contributes to the regulation of
GABAA receptor activity by voltage. Mol. Cell. Neurosci. 20, 683–694 (2002).
50. Renaud, J.-P. et al. Cryo-EM in drug discovery: achievements, limitations and
prospects. Nat. Rev. Drug Discov. 17, 471–492 (2018).
Acknowledgements We thank G. Cannone and S. Chen for cryo-EM assistance
at the MRC-LMB; Y. Chaban and K. Dent for cryo-EM assistance at the
Electron Bio-Imaging Centre (eBIC), Diamond Light Source; J. García-Nafría,
L. Dong, T. Nakane and S. Scheres for advice on cryo-EM data processing;
J. Grimmett and T. Darling for IT and computing support; and members of
the Aricescu laboratory for discussions and comments on the manuscript.
This work was supported by the UK Medical Research Council grants MR/
L009609/1, MC_UP_1201/15 (A.R.A., S.M. and D.L.) and MC_UP_A025_1013
(J.Z.); UK Biotechnology and Biological Sciences Research Council grant
BB/M024709/1 (A.R.A. and D.L.); Human Frontier Science Program grant
RGP0065/2014 (A.R.A.); Cancer Research UK grant C20724/A14414 (T.M.);
and Swiss National Science Foundation fellowship 168735 (J.Z.). R.D. and
K.W.M. were supported by a grant from the National Institute for General
Medical Sciences (GM 58448) and by the Department of Anesthesia, Critical
Care and Pain Medicine, Massachusetts General Hospital. We acknowledge
the support and the use of resources of Instruct-ERIC (PID1271), part of
the European Strategy Forum on Research Infrastructures (ESFRI), and the
Research Foundation-Flanders (FWO) for their support of nanobody discovery,
and FWO for a doctoral fellowship to T.U.
Reviewer information Nature thanks M. Chebib, M. Jansen and H. Nury for their
contribution to the peer review of this work.
Author contributions S.M. and A.R.A. conceived the project. S.M. carried out
protein purification, collected and processed the cryo-EM data, and built and
refined the models, with assistance from A.R.A. R.D. and K.W.M. designed
and analysed the electrophysiological experiments, which were performed
by R.D. T.U., E.P. and J.S. designed and generated Mb38. I.S.M. contributed to
sample screening by cryo-EM. D.L. developed Volta Phase Plate data collection
protocols and contributed to model building. D.K. and A.K. contributed to
cryo-EM data collection on the Krios-S microscope. T.M. contributed to the
analysis of structural data. J.Z. developed contrast transfer function refinement
algorithms. S.M. and A.R.A. wrote the manuscript with input from all co-authors.
Competing interests The authors declare no competing interests.
Additional information
Extended data is available for this paper at https://doi.org/10.1038/s41586-
018-0832-5.
Supplementary information is available for this paper at https://doi.org/
10.1038/s41586-018-0832-5.
Reprints and permissions information is available at http://www.nature.com/
reprints.
Correspondence and requests for materials should be addressed to S.M. or
K.W.M. or A.R.A.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

Methods
Data reporting. No statistical methods were used to predetermine sample size.
The experiments were not randomized and the investigators were not blinded to
allocation during experiments and outcome assessment.
GABAA receptor production, purification and nanodisc reconstitution.
Human tri-heteromeric α1β3γ2L was expressed in a stable, doxycycline-inducible
HEK293S-TetR cell line28 grown in suspension. The cell line was not authenticated
or tested for mycoplasma contamination. The receptor comprised the full-length
human α1 (UniProtKB P14867), β3 (P28472) and γ2L (P18507-2) subunits, each
under individual antibiotic selection (zeocin, hygromycin-B, geneticin and blastici-
din, respectively). For quality control and purification, a Flag tag (DYKDDDDK)51
was fused to the α1 subunit N terminus and a rhodopsin-1D4 tag (TETSQVAPA)52
was fused to the γ2L subunit C terminus, downstream of a (GGS)3GK spacer
sequence. Cells were grown in FreeStyle 293 expression medium (Gibco) sup-
plemented with 1% fetal bovine serum (Invitrogen), and antibiotics for selection
(geneticin, hygromycin-B, zeocin and blasticidin, Thermo Fisher Scientific) at
37 °C, 160 r.p.m., 8% CO2. GABAA receptor expression was induced by the addi-
tion of doxycycline (2 μg ml−1, Sigma) at a cell density of around 2 × 106 cells per
ml. At the same time, the medium was supplemented with 5 mM sodium butyrate
and the class I α-mannosidase inhibitor kifunensine (1 mg l−1, Toronto Research
Chemicals), in order to boost recombinant protein yields. Owing to this treatment,
the majority of N-linked glycans were restricted to the immature, endoplasmic-re-
ticulum-type Man9GlcNAc2. Cells were collected by centrifugation around 24 h
after doxycycline addition. All steps of purification were performed at 4 °C or on
ice. Cell pellets from 0.4 l of suspension culture were resuspended by brief vor-
texing in dilution buffer (50 mM HEPES pH 7.6, 300 mM NaCl) supplemented
with 1% (v/v) mammalian protease inhibitor cocktail (Sigma-Aldrich). Membrane
proteins were solubilized with 1% (w/v) n-dodecyl β-d-maltopyranoside (DDM,
Anatrace; used for BCC- and DZP/GABA-bound samples) or lauryl maltose neo-
pentyl glycol (LMNG, Anatrace; used for PTX-, PTX/GABA-, ALP/GABA-bound
samples) with cholesterol hemisuccinate (CHS, Anatrace) at a 10:1 molar ratio,
respectively, for 1 h. Insoluble material was removed by centrifugation (10,000g,
30 min). The α1β3γ2 GABAA receptors in the supernatant were captured on 1D4
affinity resin13 (300 μl) while mixing slowly for 2 h. The beads were recovered by
centrifugation (300g, 5 min) and washed with 100 ml of dilution buffer supple-
mented with 0.1% (w/v) DDM:CHS (10:1) or LMNG:CHS (10:1). To reduce the
time α1β3γ2 GABAA receptor spends solubilized in detergent, and to streamline
protein specimen preparation for cryo-EM, we reconstituted the α1β3γ2 heteromer
into nanodiscs while it was bound to the 1D4 beads. Receptors were equilibrated
with 1 ml of dilution buffer containing an excess (40 μl) of a phosphatidylcholine
(POPC, Avanti) and bovine brain lipid (BBL) extract (type I, Folch fraction I,
Sigma-Aldrich) mixture (POPC:BBL = 85:15) for 30 min. POPC and BBL extract
stocks (10 and 20 mg ml−1, respectively) were prepared by solubilization in 3%
DDM. Beads were collected by centrifugation and an excess of MSP2N2
(0.6 mg ml−1 final concentration) was added together with Bio-Beads (40 mg ml−1
final concentration) and incubated for 2 h rotating gently. The MSP2N2 belt
protein was produced as previously described53. After nanodisc reconstitution, the
1D4 resin and Bio-Bead mixture was washed extensively with dilution buffer to
remove empty nanodiscs. For protein elution, 100 μl of mixture containing 1 part
of dilution buffer and 3 parts of 2 mM 1D4 peptide stock (in MilliQ-grade H2O)
was added onto the 1D4 resin for incubation overnight. The next day, beads were
settled down by a short centrifugation (300g, 5 min) and the eluate was collected.
Typically, the eluate contained 0.1–0.25 mg ml−1 of α1β3γ2 heteromer, which was
then directly used for cryo-EM grid preparation.
Mb38 production and purification. Megabodies (Mbs) are chimeric proteins
comprising nanobodies (Nbs) fused to larger scaffold proteins (J.S., manuscript in
preparation). A circular permutant of the extracellular adhesin domain of
Helicobacter pylori (HopQ)54 was inserted into the first β-turn connecting β-strands
A and B of anti-α1 subunit Nb3820 resulting in monomeric MbcHopQ Nb38 (approxi-
mately 58 kDa). MbcHopQ
Nb38 was expressed as a C-terminally His6- and EPEA-tagged
periplasmic protein in E. coli WK6 cells. Cell cultures were grown at 37 °C (120
r.p.m.) in Terrific Broth medium supplemented with ampicillin to an optical den-
sity of 0.8 at 600 nm followed by induction with 1 mM isopropyl β-d-1-thioga-
lactopyranoside and overnight expression at 28 °C (120 r.p.m.). MbcHopQ Nb38 was
extracted from periplasm by osmotic shock55, and further purified using nickel
affinity chromatography and size-exclusion chromatography (Superdex 200 16/60
column, GE Healthcare) in 10 mM Tris pH 7.3, 140 mM NaCl at 21 °C. Purified
MbcHopQ −1
Nb38 was concentrated to around 15 mg ml , frozen in liquid nitrogen and
stored at −80 °C.
Cryo-EM sample preparation. α1β3γ2L heteromeric receptors reconstituted into
nanodiscs were supplemented with small-molecule compounds at the concen-
trations indicated in the main text, and with Mb38 (1–2 μM) to improve particle
alignment and the proportion of side views in cryo-EM images. In brief, 3.5 μl of
sample was applied onto glow-discharged gold R1.2/1.3 300 mesh UltraAuFoil
© 2018 Springer Nature Limited. All rights reserved.
Article RESEARCH
grids (Quantifoil) for 30 s and then blotted for 5.5 s before plunge-freezing the grids
into liquid ethane cooled by liquid nitrogen. Plunge-freezing was performed using
a Vitrobot Mark IV (Thermo Fisher Scientific) at approximately 100% humidity
and 14.5 °C.
Cryo-EM image collection and processing. Cryo-EM data were collected on a
300 kV Titan Krios microscopes (Thermo Fisher Scientific) fitted with a GIF-
Quantum energy filter (Gatan) and Volta Phase Plate (Thermo Fisher Scientific).
Micrographs were recorded in counting mode using K2 Summit (Gatan) or Falcon
3EC (Thermo Fisher Scientific) direct electron detectors. For sample-specific data
collection parameters, see Extended Data Table 2.
Datasets were analysed using the same basic processing pipeline in RELION
3.056 as described below. First, MotionCor257 and Gctf58 wrappers in RELION 3.0
were used to motion-correct movies, and to estimate the contrast transfer function
and phase shift parameters, respectively. Poor-quality images were discarded after
manual inspection. For particle picking, a Gaussian blob was used as a template to
auto-pick particles from a small set of micrographs and 2D classification was per-
formed. Selected 2D classes were used as templates for auto-picking particles from
all micrographs. Two rounds of reference-free 2D classifications were performed
and well-aligned 2D classes showing clear GABAA receptor projections were
selected for 3D reconstruction. An initial reference-free 3D model was generated
in RELION 3.0 using stochastic gradient descent (SGD) methodology59. Selected
particles were 3D-refined (‘gold standard’ refinement) and Bayesian polishing of
particles was performed60. Next, the polished particles were classified into eight
3D volumes without particle alignment. Particles from volumes with the highest
resolution were combined and 3D-refined using a soft mask and solvent-flattened
Fourier shell correlations (FSCs). Beam tilt correction and per-particle contrast
transfer function refinement implementations in RELION 3.0 were applied, result-
ing in further 0.1–0.2 Å increase in resolution. Notably, for the DZP/GABA-bound
GABAA receptor dataset, beam tilt correction helped to increase the resolution
from 3.82 Å to 3.58 Å. The resolution was estimated using relion_postprocess with
the FSC criteria of 0.143. Local map resolution was estimated with MonoRes61.
The initial atomic model used in this work was obtained from the truncated
α1β3γ2 heteromer structure20. First, ECDs and TMDs from this model were fit
to the PTX/GABA-bound α1β3γ2 heteromer map (3.04 Å nominal resolution)
as rigid bodies using UCSF Chimera62. Coot63 was used to adjust the model and
build the regions absent in the truncated receptor form. The model was then sub-
jected to several rounds of global refinement and minimization in real space using
phenix_real_space_refine64. The resulting model served as a starting point for the
other structures, applying the same refinement strategy. The geometry constraint
files for small-molecule ligands used in the refinement were generated using the
Grade Web Server (Global Phasing). The quality assessment of geometry in all
models was performed using the MolProbity65 Web Server. For the refinement
protocol validation, each final model coordinates were displaced by 0.5 Å and
refined using phenix_real_space_refine against one of each half-map sets pro-
duced by RELION 3.0. FSC curves were then calculated between this model and
the half-map used for refinement (‘work’) and the half-map, which was not used
for refinement, (‘free’) using phenix.mtriage66. In addition, the refined models
were also compared against the final sharpened map (‘full’). For all structures, the
separation between FSCwork and FSCfree curves was not significant, indicating that
the models were not over-refined.
Subunit interface surface and associated free energy parameter analysis was
performed using the PDBePISA server67. Pore diameters were calculated using the
HOLE68 plug-in in Coot. Structural figures were prepared using UCSF Chimera62.
Global TMD alignments were performed using the align command in PyMol
(Schrödinger). TMD boundaries for global alignments are as follows: α1 residues
233–309 and 392–418, β3 residues 218–302 and 419–447, γ2 residues 233–319 and
411–436. ECD alignments were performed using the match command in UCSF
Chimera. ECD boundaries for structural alignments: α1 residues 9–222, β3 res-
idues 8–217, γ2 residues 26–232. Root mean square deviations (r.m.s.d.) were
calculated using the r.m.s.d. function in UCSF Chimera. M2–M3 loop boundaries
for r.m.s.d. calculations are as follows: α1 residues 276–284, β3 residues 271–279,
γ2 residues 286–294. Rotation angles were calculated using UCSF Chimera.
Electrophysiology. HEK293S cells producing α1β3γ2L receptors28 were grown
on glass coverslips, and expression was induced with 0.1–2 μg ml−1 doxycycline
for 14 to 28 h depending on the level of current required. For pulling outside-out
macro-patches, we used poly-l-lysine-coated glass coverslips (Becton Dickinson).
Cells were also treated with kifunensine (5 μg ml−1) at the time of induction.
Currents were recorded in either whole-cell or outside-out configuration of
the patch clamp using an Axopatch 200A amplifier (Molecular Devices). Ligands
were applied via a quad-channel super-perfusion pipette coupled to a piezoelectric
element that switched the super-perfusion solution in <1 ms as described previ-
ously69. Data were acquired using Clampex 8.2 (Molecular Devices). For short
drug applications, data were acquired at 10 kHz and filtered at 5 kHz, whereas
for longer applications, data were collected at 2 kHz and filtered at 1 kHz. Cells or
 RESEARCH Article
patches were continually perfused with a bath solution consisting of (in mM): 145
NaCl, 5 KCl, 10 HEPES, 2 CaCl2, 1 MgCl2 and 10 glucose, pH 7.4 (adjusted with
NaOH). The pipette solution consisted of (in mM): 140 KCl, 10 HEPES, 1 EGTA, 2
MgCl2 and 2 Mg-ATP, pH 7.3 (adjusted with KOH). For whole-cell recordings, the
open pipette resistances ranged from 1.3–2.5 MΩ, and for outside-out patches they
ranged from 6 to 10 MΩ. For whole-cell recordings, cell capacitance ranged from
5 to 13 pF and series resistance ranged from 0.2 to 2.5 MΩ. Series resistances were
electronically compensated by >85% with a lag of 10 μs. Cells and patches were
voltage-clamped at –52 mV unless otherwise stated. The liquid junction potential
of –2 mV between the bath and pipette solution was corrected post-recording.
Electrophysiology data analysis. Clampfit 9 (Molecular Devices) was used to
measure the peak current amplitudes. Current traces were normalized and pre-
pared for presentation in Origin 6 (OriginLab Corporation). All data are pre-
sented as mean ± s.d. Statistical analysis was performed using Prism 6 (GraphPad
Software).
Materials. All reagents for electrophysiology solutions were purchased from either
Millipore Sigma or Thermo Fisher Scientific. Diazepam, doxycycline and DMSO
were purchased from Millipore Sigma. Bicuculline and kifunensine were purchased
from Tocris. A stock solution of bicuculline (100 mM) was prepared fresh each
day in DMSO. The stock solution of diazepam (500 mM) in DMSO was stored at
−80 °C. A working solution of diazepam (100 μM) in the bath solution was pre-
pared fresh daily and was dissolved by sonication. In control experiments, patches
exposed to GABA (10 mM) in the presence of DMSO (0.02% used in application
of diazepam) did not differ from those exposed to GABA (10 mM) alone. The
small Mb38-elicited currents were enhanced by a similar small amount by 0.1%
DMSO (84 ± 41% (n = 3)).
Reporting Summary. Further information on research design is available in
the Nature Research Reporting Summary linked to this paper.
Data availability
All data are available from the corresponding authors and/or included in the manu-
script or Supplementary Information. Atomic coordinates for PTX-, PTX/GABA-,
BCC-, ALP/GABA- and DZP/GABA-bound structures have been deposited in the
Protein Data Bank with accession codes 6HUG, 6HUJ, 6HUK, 6HUO and 6HUP,
respectively, and the cryo-EM density maps have been deposited in the Electron
Microscopy Data Bank with accession codes EMD-0275, EMD-0279, EMD-0280,
EMD-0282 and EMD-0283, respectively.
© 2018 Springer Nature Limited. All rights reserved.
51. Einhauer, A. & Jungbauer, A. The FLAG peptide, a versatile fusion tag for the
purification of recombinant proteins. J. Biochem. Biophys. Methods 49, 455–465
(2001).
52. Molday, R. S. & MacKenzie, D. Monoclonal antibodies to rhodopsin:
characterization, cross-reactivity, and application as structural probes.
Biochemistry 22, 653–660 (1983).
53. Ritchie, T. K. et al. Reconstitution of membrane proteins in phospholipid bilayer
nanodiscs. Methods Enzymol. 464, 211–231 (2009).
54. Javaheri, A. et al. Helicobacter pylori adhesin HopQ engages in a virulence-
enhancing interaction with human CEACAMs. Nat. Microbiol. 2, 16189 (2016).
55. Pardon, E. et al. A general protocol for the generation of nanobodies for
structural biology. Nat. Protoc. 9, 674–693 (2014).
56. Zivanov, J. et al. New tools for automated high-resolution cryo-EM structure
determination in RELION-3. eLife 7, e42166 (2018).
57. Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion
for improved cryo-electron microscopy. Nat. Methods 14, 331–332 (2017).
58. Zhang, K. Gctf: Real-time CTF determination and correction. J. Struct. Biol. 193,
1–12 (2016).
59. Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms
for rapid unsupervised cryo-EM structure determination. Nat. Methods 14,
290–296 (2017).
60. Zivanov, J., Nakane, T. & Scheres, S. H. W. A Bayesian approach to beam-
induced motion correction in cryo-EM single-particle analysis. IUCrJ https://doi.
org/10.1107/S205225251801463X (2019).
61. Vilas, J. L. et al. MonoRes: automatic and accurate estimation of local resolution
for electron microscopy maps. Structure 26, 337–344 (2018).
62. Pettersen, E. F. et al. UCSF Chimera—a visualization system for exploratory
research and analysis. J. Comput. Chem. 25, 1605–1612 (2004).
63. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of
Coot. Acta Crystallogr. D 66, 486–501 (2010).
64. Afonine, P. V. et al. Real-space refinement in PHENIX for cryo-EM and
crystallography. Acta Crystallogr. D 74, 531–544 (2018).
65. Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecular
crystallography. Acta Crystallogr. D 66, 12–21 (2010).
66. Afonine, P. V. et al. New tools for the analysis and validation of cryo-EM maps
and atomic models. Acta Crystallogr. D 74, 814–840 (2018).
67. Krissinel, E. & Henrick, K. Inference of macromolecular assemblies from
crystalline state. J. Mol. Biol. 372, 774–797 (2007).
68. Smart, O. S., Neduvelil, J. G., Wang, X., Wallace, B. A. & Sansom, M. S. P. HOLE: a
program for the analysis of the pore dimensions of ion channel structural
models. J. Mol. Graph. 14, 354–360, 376 (1996).
69. Forman, S. A. A hydrophobic photolabel inhibits nicotinic acetylcholine
receptors via open-channel block following a slow step. Biochemistry 38,
14559–14564 (1999).

Extended Data Fig. 1 | Single-particle cryo-EM analysis of human
α1β3γ2L GABAA receptor bound to the channel-blocker picrotoxin.
a, Representative micrograph of the PTX-bound GABAA receptor particles
embedded in vitreous ice. b, Representative 2D class averages. c, FSC
curves for the 3D reconstruction using gold-standard refinement in
RELION56. Curves are shown for the phase randomization, unmasked,
masked and phase-randomization-corrected masked maps. d, Validation
of model refinement protocol. Curves are shown for model versus summed
map (FSCfull), model refined in half-map 1 versus half-map 1 (FSCwork)
© 2018 Springer Nature Limited. All rights reserved.
Article RESEARCH
and model refined in half-map 1 versus half-map 2 (FSCfree). e, The final,
unsharpened cryo-EM map coloured by local resolution (estimated using
MonoRes61) shown at a higher contour level (left) and at a lower level
(middle and right) to highlight the nanodisc belt and flexible intracellular
domains (ICDs). f, Angular distribution of particle projections. The map
of the GABAA receptor–PTX complex is shown in teal. g, Cryo-EM density
segments for the PTX-binding site between residues 2ʹ and 9ʹ of the M2
transmembrane helices.
 RESEARCH Article

Extended Data 2 | See next page for caption.

© 2018 Springer Nature Limited. All rights reserved.


Extended Data 2 | Structural and electrophysiological analyses of
human α1β3γ2L GABAA receptor in complex with PTX and GABA.
a, FSC curves for the 3D reconstruction of the GABAA receptor bound
to PTX and GABA. Curves are shown for the phase randomization,
unmasked, masked and phase-randomization-corrected masked maps.
b, Validation of the model refinement protocol. Curves are shown for
model versus summed map (FSCfull), model refined in half-map 1 versus
half-map 1 (FSCwork) and model refined in half-map 1 versus half-map
2 (FSCfree). c, The final, unsharpened cryo-EM map coloured by local
resolution (estimated using MonoRes61). d, Superposition of the PTX-
bound and PTX/GABA-bound α1β3γ2 receptor based on the global
TMD alignment. The GABA-induced movements of loop-C in each
of the β3 subunits are highlighted by green lines between Cα atoms of
Thr202 residues. GABA is shown as spheres (carbon, khaki; nitrogen,
blue; oxygen, red). e, Cryo-EM density segments showing GABA-binding
sites in the PTX/GABA-bound structure. f–h, Representative whole-cell
current traces elicited from the same HEK293 cell by three 8.8-s pulses
© 2018 Springer Nature Limited. All rights reserved.
Article RESEARCH
of GABA (5 μM) plus Mb38 (2 μM), each separated by a 1-min wash:
control (f); one second after the start of the second 8.8-s pulse, PTX
(800 μM) was co-applied for 4 s (g); wash control showing full recovery
(h). PTX inhibited currents by 106 ± 2.6% (mean ± s.d.; n = 6 cells). In
addition, the protocol was repeated with outside-out patches (117 ± 9%
(mean ± s.d.); n = 5 patches). i, Globally superposed PTX-bound and
PTX/GABA-bound α1β3γ2 receptor transmembrane domains viewed
from the extracellular space. Side chains of 9ʹ Leu residues are shown as
sticks, whereas PTX is represented as balls and sticks. j, k, Superposition
of α1 subunit ECDs from PTX-bound and PTX/GABA-bound α1β3γ2
receptors reveal the relative β3 ECD motions towards α1 ECDs, as viewed
from outside of the receptor (j) and from the vestibule (k). Differences in
distances (Å) between the selected Cα atoms in the complexes without and
with GABA are indicated by lines. The PTX-bound structure is shown in
grey and the PTX/GABA-bound structure is coloured by subunit (α1, red;
β3, blue; γ2, yellow).
RESEARCH Article
Extended Data Fig. 3 | Structural analysis of human α1β3γ2L GABAA
receptor bound to the competitive antagonist bicuculline. a, FSC
curves for the 3D reconstruction of the GABAA receptor bound to BCC.
Curves are shown for the phase randomization, unmasked, masked and
phase-randomization-corrected masked maps. b, Validation of the model
refinement protocol. Curves are shown for model versus summed map
(FSCfull), model refined in half-map 1 versus half-map 1 (FSCwork) and
model refined in half-map 1 versus half-map 2 (FSCfree). c, The final,
unsharpened cryo-EM map coloured by local resolution (estimated using
MonoRes61). d, Cryo-EM density maps of the BCC-binding pockets.
© 2018 Springer Nature Limited. All rights reserved.
e, Superposition of the PTX-bound and BCC-bound α1β3γ2 receptor
based on the global TMD alignment. The BCC-induced movements of
loop-C in each of the β3 subunits are highlighted by green lines between
Cα atoms of Thr202 residues. BCC is shown as spheres (carbon, khaki;
nitrogen, blue; oxygen, red). f, Superposition of individual subunits from
the PTX-bound and BCC-bound GABAA receptor structures. r.m.s.d.
values (Å) for equivalent Cα in the entire subunits are shown. Loops-C
are marked by arrows. The PTX-bound structure is shown in grey and the
BCC-bound structure is coloured by subunit (α1, red; β3, blue; γ2, yellow).

Extended Data Fig. 4 | Structural analysis of human α1β3γ2L GABAA
receptor in complexes with alprazolam and diazepam. a, FSC curves
for the 3D reconstruction of the GABAA receptor bound to ALP. Curves
are shown for the phase randomization, unmasked, masked and phase-
randomization-corrected masked maps. b, Validation of the model
refinement protocol. Curves are shown for model versus summed map
(FSCfull), model refined in half-map 1 versus half-map 1 (FSCwork) and
model refined in half-map 1 versus half-map 2 (FSCfree). c, The final,
unsharpened cryo-EM map coloured by local resolution (estimated using
© 2018 Springer Nature Limited. All rights reserved.
Article RESEARCH
MonoRes61). d–f, Same as a–c but for the GABAA receptor bound to DZP.
g–i, Cryo-EM density maps of the ligand binding sites: ALP binding in
the BZD pocket (g), DZP binding in the BZD pocket (h), DZP binding
in the general-anaesthetic pocket at the β3+/α1− interface (i). Ligands
are shown as sticks, with atom colouring as follows: ALP carbons, cyan;
DZP carbons, teal; nitrogen, blue; chlorine, green. Side chains of residues
lining the binding pockets are shown as sticks and are numbered. Dotted
circles highlight the difference between the structures of alprazolam and
diazepam.
RESEARCH Article
Extended Data Fig. 5 | The classical benzodiazepines flumazenil
and bretazenil bind to the same benzodiazepine pocket in GABAA
receptors, but use different modes. ALP/GABA-bound and DZP/GABA-
bound structures are coloured by subunit (α1, red; β3, blue; γ2, yellow),
whereas the other superposed structures are shown in grey. Loop-C is
shown in a coil representation, to enable better visualization of the BZD
pocket. a, Structural formulae of flumazenil (FZL) and (S)-bretazenil
(BRZ). b, c, Superposition of γ2 subunit ECDs from the DZP-bound
α1β3γ2 and FZL-bound α1β2γ2 receptor structures reveals the FZL16
(white) position in the BZD pocket relative to DZP (teal). Side-on (b)
and top-down (c) views of the pocket are presented. d, e, Same as b, c but
© 2018 Springer Nature Limited. All rights reserved.
the structural alignment shows the relative position of the BRZ15 (white).
Grey dashed lines indicate hydrogen bonds that FZL and BRZ form with
γ2Thr142 in the BZD pocket. f, Superposition of γ2 subunit ECDs from
the PTX-, PTX/GABA-, BCC-, DZP/GABA- to ALP/GABA-bound γ2
ECD illustrates conformational changes of the BZD pocket associated with
ALP/DZP binding: an outward movement of loop-C; rearrangement of
γ2Tyr58 and γ2Phe77 side chains; and a change in the γ2Asn60 rotamer.
g, Superposition of the α1-D subunit ECDs from the PTX-bound and
ALP/GABA-bound α1β3γ2 structures shows that ALP binding causes
only a minimal outwards motion of the α1 loop-C, by 0.8 Å as measured
between Ser206 Cα atom positions.
 Article RESEARCH

Extended Data Fig. 6 | Superposition of individual subunits from r.m.s.d. values are in the range of 0.38–0.41 Å for α1 (343–344 equivalent
ALP-bound and DZP-bound α1β3γ2L GABAA receptor structures. Cα positions), 0.37–0.40 Å for β3 (334–336 equivalent Cα positions) and
Superposition of the individual subunits from the DZP-bound (grey) and 0.47 Å for γ2 subunits (330 equivalent Cα positions). Loops-C are marked
ALP-bound (α1, red; β3, blue; γ2, yellow) GABAA receptor structures. by arrows.

© 2018 Springer Nature Limited. All rights reserved.

RESEARCH Article
Extended Data Fig. 7 | Structural analysis of PTX-bound and ALP/
GABA-bound α1β3γ2L GABAA receptor structures. The PTX-bound
structure is shown in grey and the ALP/GABA-bound structure is
coloured by subunit (α1, red; β3, blue; γ2, yellow). a, b, Superposition
of α1 subunit ECDs from PTX-bound and ALP/GABA-bound GABAA
receptors reveals the relative β3 ECD motions towards α1 ECDs, as viewed
from outside of the receptor (a) and from the vestibule (b). Differences in
distances (Å) between the selected Cα atoms in the complexes without and
with GABA are indicated with lines. c, Individual subunits from the PTX-
© 2018 Springer Nature Limited. All rights reserved.
bound and the ALP/GABA-bound GABAA receptor structures superposed
on the basis of the global TMD alignment (ALP/GABA TMD over PTX
TMD). Angles between vectors representing M2 helices and the pore axis
of the PTX-bound structure are shown. Side chains of residues at the −2ʹ
and the 9ʹ positions are shown. d, Superposition of TMDs from PTX-
bound and ALP/GABA-bound structures. r.m.s.d. values (Å) are shown
for entire TMDs and for the M2–M3 loops (see Methods for boundary
definitions).

Extended Data Fig. 8 | Conformational differences at the ECD–TMD
interfaces between PTX-bound (closed) and ALP/GABA-bound
(desensitized) α1β3γ2L GABAA receptor structures. The PTX-
bound structure is shown in grey and the ALP/GABA-bound structure
is coloured by subunit (α1, red; β3, blue; γ2, yellow). The TMDs of the
principal subunits of PTX-bound and ALP/GABA-bound structures were
superposed enabling visualization of relative movements of neighbouring
ECDs and TMDs. a, b, Structural rearrangements of the ECD–TMD
interface between α1-A and β3-E subunits. c, d, Same as a, b, but for γ2-C
and β3-B subunits. e, f, Same as a, b, but for β3-B and α1-A subunits.
Amino acid residues present in the β1–β2 loop tip in each subunit are
© 2018 Springer Nature Limited. All rights reserved.
Article RESEARCH
shown, the Cαs for these residues are represented as spheres and the
distances of displacement indicated. The strictly conserved M2–M3 loop
proline residue interacting with the β1–β2 loop is shown for each subunit.
β1–β2 loop motions are indicated by curved arrows. g–l, Conformational
differences in the kM2–M3 loop and 19ʹ Arg side chain positions between
the PTX-bound and the ALP/GABA-bound structures shown for α-A
(g, h), g2-C (i, j) and b3-B (k, l) subunits. Neighbouring subunit M1
and M2 helices are shown as cylinders. Dashed lines indicate putative
hydrogen bond interactions between amino acid side chains and main-
chain carbonyls.
 RESEARCH Article

Extended Data Table 1 | Cryo-EM data collection, refinement and validation statistics

*Local resolution range.

Ψ
Resolution at which the FSC between map and model is 0.5.

© 2018 Springer Nature Limited. All rights reserved.

 Article RESEARCH

Extended Data Table 2 | Analysis of interfaces between α1β3γ2L GABAA receptor subunits

*Buried surface area per interface (sum of monomer areas buried at the interface, divided by 2, calculated using PDBePISA67).
† i
Δ G (solvation energy gain at complex formation) is the change of the solvation energy of a subunit due to interface formation, in kcal mol−1, calculated using PDBePISA67).

© 2018 Springer Nature Limited. All rights reserved.

 nature research | reporting summary
Corresponding author(s): A. Radu Aricescu

Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

Statistical parameters
When statistical analyses are reported, confirm that the following items are present in the relevant location (e.g. figure legend, table legend, main
text, or Methods section).
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
An indication of whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistics including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND
variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated

Clearly defined error bars

State explicitly what error bars represent (e.g. SD, SE, CI)

Our web collection on statistics for biologists may be useful.

Software and code

Policy information about availability of computer code
Data collection FEI EPU, Clampex 8.2

Data analysis MotionCor2, Gctf-v.1.18, RELION 2.1, RELION 3.0, MonoRes, Scipion, UCSF Chimera v1.12, Pymol v2.0.7, Coot 0.8.9, Phenix 1.13,
MolProbity, PDBePISA, HOLE, Clampfit 9.0., Graphpad Prism 6, Igor Pro, Origin 6.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers
upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
April 2018

Policy information about availability of data

All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
Atomic coordinates of all protein models were deposited in the Protein Data Bank. Cryo-EM density maps were deposited in the Electron Microscopy Data Bank.

1
 nature research | reporting summary
Field-specific reporting
Please select the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/authors/policies/ReportingSummary-flat.pdf

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.
Sample size Sample sizes were estimated on the basis of previous studies using similar methods and analyses that are widely published.

Data exclusions A small number of the acquired cryo-EM movies were discarded owing to poor ice, excessive movement or defocus.

Replication All attempts to replicate data were successful.

Randomization Randomization is not relevant to this study.

Blinding Blinding is not relevant to this study.

Reporting for specific materials, systems and methods

Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Unique biological materials ChIP-seq
Antibodies Flow cytometry
Eukaryotic cell lines MRI-based neuroimaging
Palaeontology
Animals and other organisms
Human research participants

Unique biological materials

Policy information about availability of materials
Obtaining unique materials No restrictions.

Antibodies
Antibodies used Rho 1D4 antibody was purchased from the University of British Columbia. The Mb38 megabody is available upon request.

Validation The nanobody Nb38, used to design the megabody Mb38 as described in methods, was validated by surface plasmon resonance
and published elsewhere (doi: https://doi.org/10.1101/338343).

Eukaryotic cell lines

Policy information about cell lines
April 2018

Cell line source(s) The cell line (based on ATCC CRL-3022) expressing the human alpha1beta3gamma2L GABAA receptor has been previously
described (PMID 24288268).

Authentication Authentication was not performed for this study.

Mycoplasma contamination Mycoplasma testing was not performed for this study.

2
Commonly misidentified lines N/A
(See ICLAC register)

nature research | reporting summary

April 2018